Pricing

Managed credits during the beta.

Beta workspaces receive platform-managed model credits. There is no card on file, no purchase path, and no published price — because beta pricing is not set yet.

What a credit covers.

A credit is the platform’s unit of managed AI usage. Every run separates the three things that actually cost money.

Line itemWhat it measuresUnit
Model tokensInput and output tokens for each model call in the run.tokens
Cell computeSeconds the Cell was awake. A paused Cell bills nothing.seconds
Tool and provider callsCalls routed through a connected business system.calls

Reserve, settle, release.

Credits are held before a run starts and resolved when it finishes. A run can never spend more than it reserved, and a workspace limit is enforced at the moment of the hold — not discovered afterwards.

Limits

Each workspace has a per-run ceiling so one runaway task cannot drain a balance, and an optional company ceiling on total spend. Hitting either one stops the run before any model call, with a message that says which limit was reached.

Receipts

Every settlement writes an itemized receipt: what was held, what was actually used, and what was released. Receipts and the full ledger are in your workspace and in the data export.

Bring your own key

The provider layer is designed for BYOK, where you pay the model provider directly and the platform meters only compute and tool usage. It is not enabled in the beta, so nothing here implies it is available today.

What we will not do: proxy a consumer ChatGPT or Claude subscription as API access, quote a price we have not set, or bill a Cell that is paused.

Join the 25-company beta